Struct UnifiedBuffer 

pub struct UnifiedBuffer<T: Copy> { /* private fields */ }

A contiguous buffer of T elements in CUDA unified (managed) memory.

Unified memory is accessible from both the host CPU and the GPU device. The CUDA driver transparently migrates pages between host and device as needed. This simplifies programming at the cost of potential migration overhead compared to explicit device buffers.

Implementations

impl<T: Copy> UnifiedBuffer<T>

pub fn alloc(n: usize) -> CudaResult<Self>

Allocates a unified memory buffer capable of holding n elements of type T.

The memory is allocated with CU_MEM_ATTACH_GLOBAL, making it accessible from any stream on any device in the system.

Errors

• CudaError::InvalidValue if n is zero.
• CudaError::OutOfMemory if the allocation fails.
• Other driver errors from cuMemAllocManaged.

pub fn len(&self) -> usize

Returns the number of T elements in this buffer.

pub fn is_empty(&self) -> bool

Returns true if the buffer contains zero elements.

pub fn byte_size(&self) -> usize

Returns the total size of the allocation in bytes.

pub fn as_device_ptr(&self) -> CUdeviceptr

Returns the raw CUdeviceptr handle for use in kernel launches and other device-side operations.

pub fn as_slice(&self) -> &[T]

Returns a shared slice over the buffer’s host-accessible contents.

Safety note

This is only safe to call when no GPU kernel is concurrently reading or writing this buffer. Synchronise the relevant stream or context before calling this method.

pub fn as_mut_slice(&mut self) -> &mut [T]

Returns a mutable slice over the buffer’s host-accessible contents.

Safety note

This is only safe to call when no GPU kernel is concurrently reading or writing this buffer. Synchronise the relevant stream or context before calling this method.

Trait Implementations

impl<T: Copy> Drop for UnifiedBuffer<T>

fn drop(&mut self)

Executes the destructor for this type.

impl<T: Copy + Send> Send for UnifiedBuffer<T>

impl<T: Copy + Sync> Sync for UnifiedBuffer<T>